Semantic data integration from heterogeneous, distributed data silos enables Digital Humanities research and application development employing a larger, mutually enriched and interlinked knowledge graph. However, data integration is challenging, involving aligning the data models and reconciling the concepts and named entities, such as persons and places. This paper concerns the entity reconciliation of person entities in military historical person registers for semantic data integration. A probabilistic record linkage process is presented to reconcile person references in different person registers with structured metadata into a single knowledge graph. The process was applied to reconcile three person registers of the popular semantic portal "WarSampo - Finnish World War 2 on the Semantic Web". The registers contain detailed information about some 100,000 people and are individually maintained by domain experts. This sets demands on the integration process to be automated and adaptable to changes in the registers. An evaluation of the record linkage results is promising, and provides some insight into military person register reconciliation in general.